Riva TNT2
Indeed, Riva TNT had a fill rate of 180 MTexels/sec. Although 250 MPixels/sec was originally stated. The technological process of 0.35 microns did not allow reaching those frequencies at which these characteristics would become real. There was only one way out. Do pretty much what nVidia did in 1998. Rework the existing chip and issue it in a new guise. Riva TNT2 - the new chip from nVidia was supposed to be the very remake that everyone was waiting for. So, what did nVidia offer us under the guise of Riva TNT2?
0.25 micron technology, which allows the chip to heat up less
Work at frequencies of 125/150 MHz.
Improved rendering unit, which allows to increase the speed by 10-15%.
RAMDAC 300 MHz, which allows you to work in ultra-high resolutions.
Support up to 32 Mb SDRAM / SGRAM memory.
Memory bandwidth - 2.4 Gb / sec.
AGP 4x support
As we can see, there have been significant changes. First of all, for the sake of which everything was conceived. Namely, speed! What is there cannot be taken away. Riva TNT2 in nominal mode has frequencies 1.36 times higher than those of Riva TNT. In addition, according to representatives of nVidia, each manufacturer will be free to set such frequencies as they want. Put faster memory and a high-quality cooler - set higher frequencies. Feel sorry for the money - be kind, lower the nominal.
Specifications NVIDIA Riva TNT 2
| Name | Riva TNT 2 |
| Core | NV5 |
| Process technology (µm) | 0.25 |
| Transistors (million) | 10.5~15 |
| Core frequency | 125 |
| Memory frequency (DDR) | 150 |
| Bus and memory type | SDR-128bit |
| Bandwidth (Gb/s) | 2.4 |
| Pixel pipelines | 2 |
| TMU per conveyor | 1 |
| textures per clock | 2 |
| textures per pass | ? |
| Vertex conveyors | No |
| Pixel Shaders | No |
| Vertex Shaders | No |
| Fill Rate (Mpix/s) | 250 |
| Fill Rate (Mtex/s) | 250 |
| DirectX | 6.0 |
| Anti-Aliasing (Max) | No (?) |
| Anisotropic Filtering (Max) | No (?) |
| Memory | 32MB |
| Interface | AGP4x |
| RAMDAC | 300MHz |
What do higher frequencies give us? First of all - the speed of filling. It directly depends on the frequency of the memory. And on Riva TNT2 it is 250 MTexels/sec versus 180 on Riva TNT. With multitexturing disabled, the Riva TNT2 fillrate was 125 MTexels/sec versus 90 on Riva TNT. This made it possible to disable multitexturing in games on Riva TNT2 and enjoy "honest" trilinear filtering.
Since the core speed has been increased, the maximum polygon processing speed has increased from 6 to 8 MPolys/sec. And this means that the average speed has also increased. When using 0.25 µm technology, Riva TNT2 received higher prospects for overclocking. 160/183 MHz. And this is not the limit! At such frequencies, the fillrate reaches 320 MTexels/sec. At the same frequencies, Riva TNT2 is 7-15% faster than Riva TNT.
The increased framebuffer size - 32 MB together with 300MHz RAMDAC made it possible to use resolutions up to 2048x1536. In addition, the larger video memory allowed us to work more efficiently with a large amount of textures. And the increased bandwidth of the memory bus should make this job even faster.
Support for AGP 4x was out of date. Firstly, because AGP 4x motherboards are not very common, and secondly, because the DiME function is still not fully developed. Almost all Riva TNT2 boards designed to work with AGP 4x work there in 2x mode. TV - functions remained at the same level.